Skip to content

Choosing a model

Which chat model should I use?

Pick a chat model by how costly a wrong answer would be. Use a fast, low-cost model for everyday drafting and switch to a flagship model for hard or important tasks.

For everyday drafting

Fast models — the Flash, Lite and mini tiers from Google and OpenAI, and the lower-cost GPT-5.6 tiers — are quick and cheap. They handle emails, rewrites, summaries, brainstorming and short answers well. Start here for most writing.

For hard reasoning and long work

Flagship models such as GPT-6 Astra, Claude Opus and Gemini Pro are stronger at multi-step reasoning, careful writing, long documents and tricky code. They cost more per token, so use them when a wrong answer would cost you time or money.

For coding

Claude Sonnet and Opus are widely used for code review, refactors and debugging. GPT Codex models are tuned for code as well. DeepSeek and GLM are lower-cost options for routine coding tasks and bulk changes.

For volume and low cost

DeepSeek, GLM, Kimi and Qwen models are among the lowest-cost chat models and handle large volumes of drafts, translations and data clean-up. They are also strong in Chinese.

Switching and comparing

You can switch models within Text Studio and send the same question to a second model to compare answers. Each chat model page shows its input and output rate per thousand tokens, so you can see what a long conversation will cost. The how to choose a chat model guide goes through these lanes in more detail.

FAQ

Do longer conversations cost more?
Yes. Each reply includes the conversation so far as input. Start a new chat when you change topic to keep costs down.
Which models can read images?
Most current flagship models from OpenAI, Anthropic and Google accept images. Some lighter models are text-only.
Is the answer the same as in the provider's own app?
The model is the same. Answers can differ because each app adds its own instructions and settings.

Related help articles